From the model card
Fields below are copied from the tags and counters on the HuggingFace repository facebook/sam3 at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Pipeline tag
- mask-generation
- Library
- Transformers
- Weight formats
- safetensors
- License tag
other— read the license file in the repo before relying on it- Language tags
- English (en)
- Downloads (HF counter at last fetch)
- 2,091,718
- Likes (HF counter at last fetch)
- 2,892
- Model card
- https://huggingface.co/facebook/sam3
Use cases
- Interactive image and video segmentation with point or box prompts
- Building annotation tools for computer vision datasets
- Video object tracking with sparse user prompts
- 3D scene segmentation pipelines for robotics or AR
Pros
- Handles image, video, and 3D modalities in one model
- Prompt-based interface — no fine-tuning needed for new categories
- Apache-2.0 licensed
- Builds on SAM 2's strong video tracking capabilities
Cons
- Large model size limits real-time deployment on edge hardware
- 3D segmentation quality still lags domain-specific point cloud models
- Prompt engineering required to achieve consistent results
- Limited benchmarking data available compared to SAM 2
Tags
transformerssafetensorssam3_videofeature-extractionsam3mask-generationenlicense:othereval-resultsendpoints_compatibleregion:usdeploy:azure